Extending attribute-oriented induction as a key-preserving data mining method

Muyeba, Maybin K.; Keane, John A.

Abstract

Attribute-Oriented Induction (AOI) is a set-oriented data mining technique used to discover descriptive patterns in large databases. The classical AOI method drops attributes that possess a large number of distinct values or have either no concept hierarchies, which includes keys to relational tables. This implies that the final rule (s) produced have no direct link to the tuples that form them. Therefore the discovered knowledge cannot be used to efficiently query specific data pertaining to this knowledge in a different relation to the learning relation.

This paper presents the key-preserving AOI algorithm (AOI-KP) with two implementation approaches. The order complexity of the algorithm is O (np), which is the same as for the enhanced AOI algorithm where n and p are the number of input and generalised tuples respectively. An application of the method is illustrated and prototype tool support and initial results are outlined with possible improvements.

• Presentation Conference Type: Conference Paper (published)
• Conference Name: Third European Conference, PKDD'99
• Start Date: Sep 15, 1999
• End Date: Sep 18, 1999
• Publication Date: 1999
• Deposit Date: Apr 7, 2025
• Print ISSN: 0302-9743
• Publisher: Springer Verlag
• Peer Reviewed: Peer Reviewed
• Pages: 448-455
• Series Number: 1704
• Book Title: Principles of Data Mining and Knowledge Discovery
• ISBN: 978-3-540-66490-1
• DOI: https://doi.org/10.1007/978-3-540-48247-5_57